Access to Pulmonary Rehabilitation among Medicare Beneficiaries with Chronic Obstructive Pulmonary Disease

Abstract

Rationale

Pulmonary rehabilitation (PR) remains substantially underused as a treatment modality for chronic obstructive pulmonary disease (COPD). A major barrier to the uptake of PR is the poor availability of and access to PR.

Objectives

To quantify patients’ access to PR centers in the United States.

Methods

Using the 100% Medicare population with coverage for 2018, four geodesic distance–based buffers of 10-, 15-, 25-, and 50-mi radii around the geographic centroid of each ZIP code with at least one beneficiary with COPD were created. Street addresses of PR centers across the continental United States were geocoded. We calculated the distance between the residential ZIP code centroid and the closest PR center. The proportions of individuals with at least one PR center available within the four distance buffers were calculated overall as well as in metropolitan, micropolitan, small-town, and rural areas.

Results

Of 62,930,784 Medicare beneficiaries, 10,376,949 (16.5%) had COPD. There were 1,696 PR centers across the United States, with one PR center for every 6,030 individuals with COPD. Mean distance to the nearest PR center was 12.4 (standard deviation, 16.6) mi. Overall, the proportions of individuals with COPD who had PR centers available within 10-, 15-, 25-, and 50-mi radii were 61.5%, 73.2%, 86.6%, and 97.1%, respectively. Proportions for rural areas were 11.3%, 24.3%, 53.4%, and 88.6%, respectively. Compared with those living in metropolitan areas, those living in rural areas were 95% less likely to have PR centers within 10 mi of their residences (odds ratio, 0.048 [95% confidence interval, 0.039–0.057]).

Conclusions

In a nationally representative sample of Medicare beneficiaries, we found that two-fifths of adults with COPD overall, and eight in nine of those in rural areas, have poor access to PR.

Chronic obstructive pulmonary disease (COPD) is the third leading cause of disability in the United States and affects approximately 24 million adults (1). Pulmonary rehabilitation (PR) is an effective therapy for alleviating dyspnea and improving quality of life and exercise capacity of individuals with COPD (2). PR is also associated with a significant reduction in hospitalization (3, 4), and initiation of PR soon after hospitalization for an acute exacerbation is associated with improved survival (5). Despite these proven benefits, PR as a treatment modality remains substantially underused, and its use has been stagnant. Between 2003 and 2012, PR use in older adults with COPD increased only marginally from 2.6% to 3.7% (6). In 2012, the rate of PR use after hospitalization was only 1.9% within 6 months and 2.7% within 12 months (7).

The reasons underlying this poor use are multifactorial and include socioeconomic disadvantages and factors related to disease morbidity (8, 9). A major barrier to uptake of PR has been the poor availability of and access to PR, a gap in care that was highlighted in the National Heart, Lung, and Blood Institute’s COPD National Action Plan (10); yet data quantifying this problem are limited. Significant geographic disparities in the location of PR centers have been previously highlighted. Moscovice and colleagues found that 56% of counties in the United States do not have PR centers, and this worsens to 73% in noncore rural counties (11). This lack of availability directly translates to a significant barrier to use. Indeed, those living more than 10 mi from PR facilities are less than half as likely to initiate PR than those living closer (7), and those with transit distances of more than 30 minutes are half as likely to complete PR than those with shorter transit distances (12). Although the lack of availability of PR centers by location has been described, the number of individuals who are disadvantaged by distance to PR centers has not been previously quantified. Whether the county-level disparities translate to disadvantages on the basis of population density is not known. Using a representative sample of Centers for Medicare and Medicaid Services (CMS) beneficiaries across the continental United States, we aimed to quantify access to PR in the United States.

Methods

Study Population

We conducted a cross-sectional analysis of Medicare beneficiaries with diagnoses of COPD using the CMS Chronic Conditions Summary and Beneficiary Summary files for the 100% Medicare population with coverage for 2018. International Classification of Diseases (ICD), Ninth or Tenth Revision, Clinical Modification codes were used to define COPD. Individuals were considered to have COPD if they had either one inpatient or two outpatient claims with ICD, Ninth Revision, diagnosis code 490, 491.x, 492.x, 494.x, or 496 or ICD, Tenth Revision, diagnosis code J40, J41.x, J42, J43.x, J44.x, or J47.x any time between January 1, 1999 and December 31, 2018. Demographics and ZIP codes of residence were recorded. As street addresses are not available for Medicare beneficiaries, we used the geographic centroids of the ZIP codes of the beneficiaries’ residences to define their residences. Four geodesic distance–based buffers of 10-, 15-, 25-, and 50-mi radii were created around the geographic centroid of each ZIP code using ArcGIS software (Esri).

PR Centers

Ambulatory PR centers, both hospital and community based, were identified in 2020 using a combination of the American Association of Cardiovascular and Pulmonary Rehabilitation directory, state rehabilitation society listings, and e-mails to academic center contacts in each state. Google Maps searches using the search term “pulmonary rehabilitation” were also used to identify additional centers. Each identified site was confirmed to be in operation by either assessing active websites or via phone calls to the centers. We excluded inpatient PR facilities. Street addresses of PR centers across the continental United States as well as Alaska and Hawaii were geocoded by assigning latitude and longitude using R statistical software version 4.0.3 (R Core Team).

Access to PR Centers

The distance between the geographic centroid of each residential ZIP code and the closest PR center was calculated using ArcMap 10.8 of ArcGIS. ArcGIS is geographic information system software developed by Esri that is widely used to process and analyze geographic data through layers in geographic maps (13, 14). ArcMap is the central application of ArcGIS, with which one can process and analyze geographic information system data sets. Using a projected coordinate system in ArcMap, we spatially joined the shapefile containing the PR center coordinates and the shapefile containing the geographic centroid of the Medicare beneficiaries’ residential ZIP codes. We used ArcMap’s “Proximity toolset,” which makes it possible to identify features that are closest to one another and calculate the distances between them or around them as buffers. Using this, we calculated the geodesic distance between the nearest PR center and the ZIP code centroid in miles (Figure 1). Geodesic distance, unlike planar distance, takes the earth’s elliptical shape into consideration while estimating distances. The proportions of beneficiaries who had at least one PR center available within the four distance buffers (10, 15, 25, and 50 mi) were calculated. Poor access to PR centers was defined as the absence of at least one PR center within a 10-mi distance from the ZIP code centroid. ZIP codes were classified using the 2010 rural-urban commuting area (RUCA) codes into metropolitan (contains an urban area of ⩾50,000 population), micropolitan (contains an urban cluster of population of 10,000–50,000), small-town (includes areas with a population <10,000, whose residents mostly commute to an urban cluster), and rural (residents’ primary commute is limited to places outside an urban area or urban cluster) areas (15). We also grouped ZIP codes by census regions (Midwest, Northeast, South, and West) and calculated access to PR as described above in regions stratified using the RUCA designation and census regions (15). Because population within a ZIP code may be concentrated in certain areas, we performed sensitivity analyses by repeating the above steps using population-based centroids for residential ZIP codes instead of their geographic centroids. This study was approved by the University of Alabama at Birmingham institutional review board, which granted a waiver of the requirement to obtain informed consent.

Figure 1.

Geodesic buffers around the centroid of a ZIP code.

Results

Of 62,930,784 Medicare beneficiaries, 10,376,949 (16.5%) had diagnoses of COPD. After excluding those living in the outlying U.S. islands and those with missing ZIP code information, we included 10,185,080 (98.2%) individuals with COPD in the spatial analyses. The average age of those with COPD was 74.3 (standard deviation [SD], 11.7) years, and 5,683,712 (55.8%) were women. The cohort was composed of 8,159,529 (80.1%) White, 1,046,668 (10.3%) Black, 597,783 (5.9%) Hispanic, 198,296 (2.0%) Asian/Pacific Islander, and 64,324 (0.6%) American Indian/Alaskan Native individuals, as well as 57,235 (0.6%) identified as “other.” Three-fourths (74.8%) of those with COPD lived in metropolitan areas, 13.1% in micropolitan areas, and 12.0% in small-town/rural areas. The majority of individuals with COPD resided in the South (n = 4,332,198 [42.5%]), followed by the Midwest (n = 2,447,259 [24.0%]), Northeast (n = 1,761,504 [17.3%]), and West (n = 1,644,088 [16.1%]).

We identified 1,696 unique PR centers, of which 1,691 were in the continental United States (Figure 2). Of these, 1,684 had identifiable RUCA codes. Six hundred forty-seven (38.2%) were located in the Midwest, 525 (31.0%) in the South, 276 (16.3%) in the Northeast, and 248 (14.6%) in West. One thousand eighty-seven (64.6%) PR centers were located in metropolitan areas, 301 (17.9%) in micropolitan areas, 223 (13.2%) in small towns, and 73 (4.3%) in rural areas. Overall, one PR center existed for every 6,030 adults with COPD. One PR center existed for every 6,980, 4,445, 3,303, and 6,030 adults with COPD in metropolitan, micropolitan, small-town, and rural areas, respectively. Participant characteristics on the basis of distance to the nearest PR center are presented in Table 1. The mean distance between the participants’ residences and the nearest PR center was 12.4 (SD, 16.6) mi; corresponding distances in metropolitan, micropolitan, small-town, and rural areas were 8.6 (SD, 10.8), 20.1 (SD, 21.2), 25.7 (SD, 23.7) and 29.4 (SD, 28.9) mi, respectively. Figure 3 is a choropleth map of the United States displaying distances between ZIP code centroids and the nearest PR center.

Figure 2.

Pulmonary rehabilitation (PR) centers in the United States, 2020. Each red dot denotes a unique PR center.

Table 1.

Characteristics of Medicare beneficiaries with chronic obstructive pulmonary disease across the United States stratified by the distance to the nearest pulmonary rehabilitation center, 2018

	0–10 mi	0–15 mi	0–25 mi	0–50 mi	>50 mi
Access to at least one PR center within the radius, n (%)	6,264,346 (61.5)	7,457,817 (73.2)	8,817,362 (86.6)	9,890,061 (97.1)	295,019 (2.9)
Age, yr, mean (SD)	74.7 (11.8)	74.6 (11.8)	74.5 (11.8)	73.3 (11.7)	74.3 (11.3)
Women, n (%)*	3,557,683 (56.8)	4,209,631 (56.5)	4,945,628 (56.1)	5,524,522 (55.9)	159,190 (54.0)
Race/ethnicity, n (%)*
White	4,800,702 (76.6)	5,829,066 (78.2)	7,028,153 (79.7)	7,939,326 (80.3)	220,203 (74.6)
Black	782,444 (12.5)	855,436 (11.5)	943,889 (10.7)	1,028,916 (10.4)	17,752 (6.0)
Hispanic	403,465 (6.4)	465,278 (6.2)	509,714 (5.8)	558,302 (5.7)	39,481 (13.4)
American Indian	20,799 (0.3)	25,767 (0.4)	34,771 (0.4)	49,859 (0.5)	14,465 (4.9)
Asian/Pacific Islander	173,502 (2.8)	185,887 (2.5)	192,149 (2.2)	197,128 (2.0)	1,168 (0.4)
Other	41,017 (0.7)	46,888 (0.6)	52,480 (0.6)	56,275 (0.6)	960 (0.3)
Unknown	42,417 (0.7)	49,495 (0.7)	56,206 (0.6)	60,255 (0.6)	990 (0.3)
Rurality, n (%)*
Metropolitan	5,540,945 (72.8)	6,460,805 (84.9)	7,183,656 (94.4)	7,541,518 (99.1)	66,776 (0.9)
Micropolitan	511,719 (38.1)	646,792 (48.2)	937,875 (69.9)	1,244,527 (92.7)	97,765 (7.3)
Small town	155,488 (21.0)	230,032 (31.1)	431,537 (58.3)	665,566 (90.0)	74,247 (10.0)
Rural	55,959 (11.3)	119,914 (24.3)	263,944 (53.4)	438,043 (88.6)	56,213 (11.4)
Geographic region, n (%)*
Midwest	1,673,280 (68.4)	1,941,150 (79.3)	2,231,030 (91.2)	2,428,966 (99.3)	18,293 (0.8)
Northeast	1,289,697 (73.2)	1,490,777 (84.6)	1,679,427 (95.3)	1,756,249 (99.7)	5,255 (0.3)
South	2,243,927 (51.8)	2,813,068 (64.9)	3,550,635 (82.0)	4,169,475 (96.2)	162,723 (3.8)
West	1,057,428 (64.3)	1,212,797 (73.8)	1,356,241 (82.5)	1,535,340 (93.4)	108,748 (6.6)

Definition of abbreviation: PR = pulmonary rehabilitation; SD = standard deviation.

^*Percentage within a buffer for each geographic region.

Figure 3.

Availability of pulmonary rehabilitation (PR) centers among Medicare beneficiaries with chronic obstructive pulmonary disease across the United States, 2018. PR center information is for 2020, and Medicare data are for 2018.

Overall, 61.5% of individuals with COPD had at least one PR center within 10 mi of their residences (Table 1). The proportions of individuals with PR centers within 15-, 25-, and 50-mi radius were 73.2%, 86.7%, and 97.1%, respectively. Access was generally better in metropolitan areas (72.8%, 84.9%, 94.4%, and 99.1% within 10-, 15-, 25-, and 50-mi buffers, respectively) and worse in rural areas (11.3%, 24.3%, 53.4%, and 88.6% within 10-, 15-, 25-, and 50-mi radii, respectively). Compared with those living in metropolitan areas, those living in rural areas were 95% less likely to have PR centers within 10 mi of their residences (odds ratio, 0.048 [95% confidence interval, 0.039–0.057]) (Table 2).

Table 2.

Access to pulmonary rehabilitation centers among Medicare beneficiaries with chronic obstructive pulmonary disease by geographic region and rurality

	Odds Ratio	95% Confidence Interval
Geographic region
Northeast	Reference	—
Midwest	0.791	0.787–0.795
West	0.659	0.657–0.663
South	0.393	0.392–0.395
Rurality
Metropolitan	Reference	—
Micropolitan	0.230	0.226–0.231
Small town	0.099	0.0987–0.0999
Rural	0.048	0.039–0.057
Rurality within geographic regions
Northeast
Metropolitan	Reference	—
Micropolitan	0.170	0.168–0.172
Small town	0.144	0.142–0.147
Rural	0.046	0.045–0.047
Midwest
Metropolitan	Reference	—
Micropolitan	0.215	0.213–0.216
Small town	0.085	0.0838–0.0854
Rural	0.030	0.0298–0.0307
West
Metropolitan	Reference	—
Micropolitan	0.150	0.148–0.151
Small town	0.049	0.048–0.050
Rural	0.011	0.010–0.0114
South
Metropolitan	Reference	—
Micropolitan	0.268	0.266–0.269
Small town	0.098	0.097–0.099
Rural	0.071	0.070–0.072

In comparisons by geographic regions, individuals in the southern United States had the lowest access to PR. Only 51.8% of those in the South had at least one PR center within 10 mi, compared with 64.3% in the West, 68.4% in the Midwest, and 73.2% in the Northeast (Table 1). Compared with those living in the Northeast, where access was the best, those living in the South were 61% less likely to have PR centers within 10 mi of their residences (odds ratio, 0.393 [95% confidence interval, 0.392–0.395]) (Table 2). Given the differences in population density across geographic regions, we conducted additional analyses comparing access across the spectrum of rurality within each geographic region. The poor access in rural areas persisted across all geographic areas, including in the northeastern United States, where PR access was the best by region (Table 2).

In sensitivity analyses using population-based centroids, the results were similar. The total number of beneficiaries included in this analysis was 9,852,391 (94.9% of total beneficiaries with COPD) because of some missing population-based centroids. The mean distance to the nearest PR center was 12.2 (SD, 16.3) mi overall and 8.5 (SD, 10.6), 19.6 (SD, 20.7), 25.3 (SD, 23.5), and 29 (SD, 28.5) mi in metropolitan, micropolitan, small-town and rural areas, respectively. Cumulatively, the proportions of beneficiaries with COPD who had at least one PR center available within 10-, 15-, 25-, and 50-mi radii of population-weighted centroids were 61.6%, 73.4%, 86.6%, and 97.1%. Individuals in rural areas had poorer access to PR than those in micropolitan or metropolitan areas; only 11.5% of rural residents had at least one PR center within 10 mi, compared with 21.1% in small-town, 38.5% in micropolitan, and 72.7% in metropolitan areas.

Discussion

In a nationally representative sample of Medicare beneficiaries, we found that two-fifths of older adults with COPD in the United States have poor access to PR. The lack of access to PR is more pronounced in rural areas, where only one in nine had PR centers available within 10 mi. Although the southern United States has the largest number of individuals with COPD, the lack of access to PR is greatest in the South.

There is considerable heterogeneity in PR availability in the United States (11, 16). Prior studies have quantified the availability of PR centers by geographic region. Spitzer and colleagues found that there is a median of four PR programs per hospital referral region (HRR), with a median PR program density of only 0.06 for every 1,000 hospitalized Medicare recipients (16). Moscovice and colleagues found that the proportion of hospitals in the United States that offer outpatient PR varies by state from 4.6% to 85.7% and that approximately three-fourths of rural counties do not have PR centers (11). These two studies were limited by how regions were determined. HRRs are large regions demarcated for the purpose of defining local healthcare markets, and the entire country is divided into only 306 HRRs. Counties are administrative divisions set on the basis of convenient geographic demarcations, and both the population and the area of each county vary considerably. Given that PR centers are more likely to be located in densely populated urban areas, our results that are based on distance from a patient’s perspective provide additional clarity about the magnitude of access issues. In addition, we included all patients with diagnoses of COPD, in contrast to prior studies that included only hospitalized patients.

Our finding that two-fifths of adults with COPD overall, and eight of nine in rural areas, do not have good access to PR is likely an underestimate. Overall, there is only one PR center for every 6,030 individuals with COPD on Medicare, and this ratio is likely to be lower when individuals with COPD who are not on Medicare are included. About half of patients with COPD report being on Medicare, and the remainder receive Medicaid or have private insurance (17). It should also be noted that the presence of a PR center alone does not guarantee capacity; the sizes of PR programs vary considerably, and wait times may vary accordingly, even in metropolitan areas.

There are several potential reasons for the poor availability of PR centers in the United States. A significant contributor is the poor reimbursement for PR (18). A marked reduction in Medicare reimbursement for PR in 2010 has likely led to a decline in the number of available centers, a trend accelerated by the coronavirus disease (COVID-19) pandemic (19). Each PR session is reimbursed at approximately half the rate of cardiac rehabilitation, despite the generally higher complexity of PR (19). We note that the number of PR centers identified in the study by Moscovice and colleagues differ slightly from our estimate (11). This may be due to the inclusion of hospital-based PR centers only in that study, whereas we included community-based PR centers. Many centers have shut down in the interval period between the two studies and are no longer offering PR. There is a significant lack of awareness of the benefits of PR among not just patients but also healthcare professionals (20). Although the focus of this study was on COPD, access to PR for those with other chronic respiratory diseases is also likely to be as poor and needs evaluation. Our data highlight the need for a concerted plan to improve PR services in the country. These include, but are not limited to, increasing reimbursements for PR services, improving awareness of PR, and implementing alternative and complementary avenues for PR, such as home PR and telehealth PR (21–24).

Strengths and Limitations

This study has a few limitations. First, the COPD diagnosis in the CMS Chronic Conditions Summary file also included those with bronchiectasis, but this constitutes a small proportion of the population. We did not have spirometric confirmation of airflow obstruction. Use of administrative data may also result in under-, over-, and misdiagnosis of COPD.

Second, although we did not include patients with COPD with other forms of insurance coverage, these data include the majority of U.S. adults with COPD, as we used the 100% Medicare population data. Third, from a payer perspective, our estimates of eligible patients with COPD may be considered overestimates. CMS covers PR for patients with moderate to very severe COPD under Medicare Part B, a recommendation mostly followed by private insurers as well. However, we and others have shown that individuals with COPD at all stages of disease severity benefit from PR (25), and hence our results are generalizable.

Fourth, we used distance as the metric for creating our buffers. We acknowledge that the time taken to drive those distances may vary by geographic location. However, travel time is difficult to quantify because of traffic and time of the day, and hence we used the more accurately quantifiable distance metric on the basis of prior studies that have documented that a 10-mi distance more than halves the odds of enrolling in PR (7).

Fifth, although we used multiple data sources to identify hospital-based PR programs, we may have misclassified some hospitals or missed some programs because of missing or outdated information on hospital websites or from our state contacts. Sixth, the Medicare database does not contain the actual addresses of beneficiaries’ homes, so we used their ZIP code centroids as a proxy. Although this is not ideal, it is comparatively a smaller spatial unit than counties and HRRs, which were used in previous studies.

Seventh, although we checked the websites of all PR facilities included, and made phone calls when websites were not available or were incomplete, it is possible that some sites may not have offered all components of PR as advertised. However, this aspect adds to the concerns of limited access to optimal PR for individuals with COPD.

Our study also has some strengths. We used a multiple-step approach to verify and include a comprehensive list of PR centers across the United States. To our knowledge, this is the first study to characterize and quantify the availability of community- and hospital-based PR centers in relation to the residential location of COPD patients.

Conclusions

Our findings of poor access to PR for patients with COPD, especially in rural areas, have important implications for improving COPD care in the United States. PR is arguably the most effective therapy for COPD, and it is imperative that we improve access for patients with COPD to this proven treatment modality.